beyond first order logic ii - university of illinois at ...homepages.math.uic.edu › ~jbaldwin ›...

BEYOND FIRST ORDER LOGIC II

John T. BaldwinDepartment of Mathematics, Statistics and Computer Science

University of Illinois at Chicago

August 8, 2004

1

WHY GO BEYOND FIRST ORDER LOGIC?

I. Because it’s there.

A. To understand the infinite:

B. To understand canonical structures

II. To understand first order logic

III. To understand ‘Model Theory’

IV. To investigate ordinary mathematical structures

2

CONTEXT

This is the second of two talks with different emphases

discussing some current work in nonelementary contexts.

This one will focus more on IB and IV. But we describe the

interplay between the ‘pure’ and ‘applied’ considerations.

3

TO UNDERSTAND THE INFINITE!

Most known mathematical results are either

extremely cardinal dependent: about finite or countable

structures or at most structures of cardinality the contin-

uum;

or completely cardinal independent: about every struc-

ture satisfying certain properties.

4

Understanding Classes of Models

Model theory has discovered problems that have an in-

timate relation between the cardinality of structures and

algebraic properties of the structures:

i) Stability spectrum and counting models

ii) A general theory of independence

iii) Decomposition theorems for general models

There are structural algebraic, not merely combinatorial

features, which are non-trivially cardinal dependent.

5

TO UNDERSTAND CANONICAL STRUCTURES

A Thesis of Zilber:

Fundamentally important structures like the complex field

wih exponentiation can be described at least up to cate-

goricity in power in an appropriate logic.

6

TO UNDERSTAND FIRST ORDER LOGIC

The study of first order logic uses without thinking such

methods as:

1. compactness theorem

2. upward and downward Lowenheim-Skolem theorem

3. closure under unions of Elementary Chains

4. Ehrenfeucht-Mostowski models

We can better understand these methods and their use in

the first order case by investigating situations where only

some of them hold.

7

TO UNDERSTAND MODEL THEORY

A. What are the syntactial and semantical components

of model theory?

B. Logics vrs classes of models: Robinson, Tarski, Morley,

Shelah

C. How does the ability to change vocabulary distinguish

model theory from other mathematical disciplines?

For applications, it is often essential to work in a given

language where one has control of the definable sets.

For general results, it is often convenient (essential?) to

expand the language to make all relations which are defin-

able, definable in a very simple way. This method is much

more powerful in the non-elementary context.

8

INVESTIGATE ORDINARY MATHEMATICS

The following classes require going beyond countable the-

ories in first order logic:

1. Banach Spaces (Krivine, Stern, Henson, Iovino, Bern-

stein, Usvyatsov, et al)

2. Complex Exponentiation (Zilber)

3. Locally finite groups (Grossberg, Macintyre, Shelah)

4. Compact complex manifolds (Zilber, Pillay, Scanlon,

Moosa, Radin)

In extending the study of finding definable groups to the

non-elementary context, Hytinnen-Lessmann-Shelah have

discovered interesting aspects of Group Representations.

9

TWO DIRECTIONS

A. Strong Syntax:

Study categoricity for sentences of Lω1,ω or Lω1,ω(Q) with

no model theoretic assumptions like upwards Lowenheim-

Skolem or the amalgamation property.

B. AEC’s with arbitrarily large models:

Study very abstract classes with some strong hypotheses.

We focus on A in this talk but make a few comments about

B first.

10

THE CATEGORICITY SPECTRUM

Theorem 1 (Morley) A countable first order theory

T is categorical in one uncountable cardinal if and only

if it is categorical in all uncountable cardinalities.

Is first order crucial? Shelah showed: ‘countable’ is not.

The study of compact complex manifolds naturally uses

uncountable languages.

1. Moosa identified Kahler manifolds as those which ad-

mit a countable language.

2. Radin generalized to certain Zariski structures in an

uncountable language properties which hold in arbi-

trary categorical theories only when the language is

countable.

Does the theorem generalize to other classes of models?

11

TWO COUNTEREXAMPLES

Let the vocabulary contain a unary predicate P .

1) L(Q) can say both the set and its compliment are un-

countable. This theory is categorical in ℵ1 and nowhere

else.

2) With an additional binary relation we can say

2|P (M)| ≥ |M |.The class of reducts is categorical in κ only if κ = iα for

some limit ordinal α.

What kinds of classes do we mean?

12

PCΓ CLASSES

A class K of τ -structures is called PC if it is the collection

of reducts to τ of the models of a first order theory T ′ in

some τ ′ ⊇ τ .

A class K of τ -structures is called PCΓ if it the collection

of reducts to τ of the models of a first order theory T ′ in

some τ ′ ⊇ τ which omit all types in a specified collection

Γ of types in finitely many variables over the empty set.

13

ABSTRACT ELEMENTARY CLASSES

Definition 2 A class of L-structures, (K,≤), is said

to be an abstract elementary class: AEC if both K and

the binary relation ≤ are closed under isomorphism

and satisfy the following conditions.

• A1. If M ≤ N then M ⊆ N .

• A2. ≤ is a partial order on K.

• A3. If 〈Ai : i < δ〉 is ≤-increasing chain:

1.⋃

i<δ Ai ∈ K;

2. for each j < δ, Aj ≤⋃

i<δ Ai

3. if each Ai ≤ M ∈ K then⋃i<δ Ai ≤ M .

• A4. If A,B,C ∈ K, A ≤ C, B ≤ C and A ⊆ B

then A ≤ B.

• A5. There is a Lowenheim-Skolem number LS(K)

such that if A ⊆ B ∈ K there is a A′ ∈ K with

A ⊆ A′ ≤ B and

|A′| < LS(K) + |A|.

14

THE PRESENTATION THEOREM

Theorem 3 Every AEC is a PCΓ.

H(κ), the Hanf number for AEC’s in vocabularies of size

κ, is the least cardinal such that:

Every AEC in vocabulary of size κ with a model of size

H(κ) has arbitrarily large models.

The presentation theorem implies:

H(κ) = iω1((2κ)+) = H1.

i.e. the Hanf number for omitting types in PCΓ classes.

H2 = H(H1).

We also get Ehrenfeucht-Mostowski models and omitting

types theorems in AEC if there is a model at least H1.

15

AMALGAMATION AND GALOIS TYPES

If the AEC (K,≤) has the amalgamation property, then

a monster model M exists. One can define:

The Galois type of a over M ∈ K is the orbit of a in Munder automorphisms of M fixing M .

Theorem 4 Model-homogenous and universal for ≤ is

the same as Galois saturated.

16

TAMENESS

In general, Galois equivalence refines having the same syn-

tactic type even in AEC’s where syntactic types make

sense.

Definition 5 K is (χ, χ1)-tame if two distinct Galois

types over a model of cardinality χ, have different re-

strictions to a submodel of cardinality χ1.

Here, Galois types are defined only over models. Recently,

Hytinnen has some very interesting ideas for defining types

over arbitrary sets in any AEC.

Any superstable theory with OTOP is ℵ0-tame but not

excellent.

17

CONTEXT

Conjecture: Let X be the class of cardinals in which a

reasonably defined class is categorical.

Not both X and the complement of X are cofinal.

(Note: So, PC-classes are not ‘reasonable’.)

We know this conjecture for first order theories and for

excellent classes in Lω1,ω. But is open even for general

sentences in Lω1,ω. So it is reasonable to investigate it first

with quite strong hypotheses.

Of course, it is only interesting when K has arbitrarily

large models – EM methods are applicable.

18

EVENTUAL CATEGORICITY

Theorem 6 (Shelah) Assume the AEC K has

1. ap and jep

2. is categorical in a successor cardinal λ > H2

A. for some χ < H(τ ) and any χ1 < λ, K is (χ1, χ)-

tame and

B. K is categorical in every µ with H2 ≤ µ ≤ λ.

Note we get (χ1, χ)-tame for small χ1 and smaller but not

minute χ. I.e. χ ∼ H1.

19

Theorem 7 (Grossberg-VanDieren) If we add the hy-

pothesis:

for some χ < H(τ ) and ANY χ1, K is (χ1, χ)-tame

then if K is categorical in two successive cardinals

above χ1, then K is categorical in all larger cardinals.

In particular, if K is categorical in one successor car-

dinal > H2, K is categorical in every θ with θ ≥ H2.

Jep is assumed for convenience.

AP is a very significant assumption

20

Categorical Structures

I. (C, =)

IIa. (C, +, =)

IIb. (C,×, =)

not quite vector spaces over Q.

III. (C, +,×, =)

Algebraically closed fields - Steinitz

IV. (C, +,×, =, exp)

Clearly not first order categorical. But maybe the only

obstruction is (Z, +,×, =).

21

GEOMETRIES

Definition. A pregeometry is a set G together with a

dependence relation

cl : P(G) → P(G)

satisfying the following axioms.

A1. cl(X) =⋃{cl(X ′) : X ′ ⊆fin X}

A2. X ⊆ cl(X)

A3. cl(cl(X)) = cl(X)

A4. If a ∈ cl(Xb) and a 6∈ cl(X), then b ∈ cl(Xa).

If points are closed the structure is called a geometry.

22

STRONGLY MINIMAL I

M is strongly minimal if every first order definable subset

of any elementary extension M ′ of M is finite or cofinite.

a ∈ acl(X) if there is a first order formula with finitely

solutions over X which is satisfied by a.

Exercise: If f takes X to Y is an elementary isomor-

phism, f extends to an elementary isomorphism from acl(X)

to acl(Y).

23

STRONGLY MINIMAL II

Lemma. A complete theory T is strongly minimal if and

only if it has infinite models and

1. algebraic closure induces a pregeometry on models of

T ;

2. any bijection between acl-bases for models of T ex-

tends to an isomorphism of the models

Theorem. A strongly minimal theory is categorical in

any uncountable cardinality.

24

QUASIMINIMALITY I

Trial Definition M is ‘quasiminimal’ if every first or-

der (Lω1,ω?) definable subset of M is countable or co-

countable.

a ∈ acl′(X) if there is a first order formula with countably

solutions over X which is satisfied by a.

Exercise ? If f takes X to Y is an elementary iso-

morphism, f extends to an elementary isomorphism from

acl′(X) to acl′(Y).

25

QUASIMINIMAL EXCELLENCE

A class (K, cl) is quasiminimal excellent if it admits a

combinatorial geometry which satisfies on each M ∈ K

1. there is a unique type of a basis,

2. a technical homogeneity condition:

ℵ0-homogeneity over ∅ and over models.

3. and the ‘excellence condition’ which follows.

If (K, cl) is quasiminimal excellence then cl is given by

the trial definition on the previous slide.

The ω-homogeneity yields by an easy induction:

Lemma 8 If cl(X) and cl(Y ) are countable then an

isomorphism between X and Y extends to an isomor-

phism of cl(X) and cl(Y )

26

QUASIMINIMAL EXCELLENCE

In the following definition it is essential that ⊂ be under-

stood as proper subset.

Definition 9 1. For any Y , cl−(Y ) =⋃

X⊂Y cl(X).

2. We call C (the union of) an n-dimensional indepen-

dent system if C = cl−(Z) and Z is an independent

set of cardinality n.

To visualize a 3-dimensional independent system think of

a cube with the empty set at one corner A and each of the

independent elements z0, z1, z2 at the corners connected

to A. Then each of cl(zi, zj) for i < j < 3 determines a

side of the cube: cl−(Z) is the union of these three sides;

cl(Z) is the entire cube.

27

EXCELLENCE

The class (K, cl) is quasiminimal excellent if it satisfies

the following additional condition.

Let C = cl−(Z) be an n-dimensional independent system.

For any a ∈ cl(Z), there is a finite X ⊂ C:

tp(a/X) |= tp(a/C).

28

EXCELLENCE IMPLIES CATEGORICITY

Theorem 10 Let K be a quasiminimal excellent class

and suppose H, H ′ ∈ K satisfy the countable closure

condition.

Let A,A′ be cl-independent subsets of H, H ′ with cl(A) =

H, cl(A′) = H ′ respectively and ψ a bijection between

A and A′.Then ψ extends to an isomorphism of H and H ′.

Thus K is categorical in every uncountable cardinality.

29

EXCELLENCE IMPLIES CATEGORICITY: PROOF

Fix a countable subset A0 of A; without loss of gener-

ality, we can assume ψ is the identity on A0 and work

over G = cl(A0). So from now on monomorphism means

monomorphism over G and cl(X) means cl(A0X).

Note that ψ is a monomorphism and so is ψ0 = ψ|A0. By

Lemma 8 and induction, for any X with |X| ≤ ℵ1, ψ|Xextends to a isomorphism from cl(X) to cl(X).

30

H = limX⊂A;|X|<ℵ0 cl(X).

The theorem follows if:

for each finite X we can choose ψX : cl(X) → H ′ so that

X ⊂ Y implies ψX ⊂ ψY .

We prove this by induction on |X|. Suppose |Y | = n + 1

and we have appropriate ψX for |X| < n + 1. We will

prove two statements by induction.

1. ψ−Y : cl−(Y ) → H ′ defined by ψ−Y =⋃

X⊂Y ψX is a

monomorphism.

2. ψ−Y extends to ψY defined on cl(Y ).

The first step is done by induction and ω-homogeneity

using Lemma 8. The exchange axiom is used to guarantee

that the maps ψ′Y for Y ′ ⊃ Y agree where more than one is

defined. The second is follows by Excellence and induction

using Lemma 8 and the fact that cl(Y ) is countable.

31

CATEGORICITY

Theorem 11 Suppose the quasiminimal excellent (I-

IV) class K is axiomatized by a sentence Σ of Lω1,ω,

and the relations y ∈ cl(x1, . . . xn) are Lω1,ω-definable.

Then, for any infinite κ there is a unique structure in

K of cardinality κ which satisfies the countable closure

property.

NOTE BENE: The categorical class could be axiomatized

in Lω1,ω(Q). But, the categoricity result does not depend

on any such axiomatization.

32

COVERS OF THE MULTIPLICATIVE GROUP OF C

Consider a short exact sequence:

0 → Z → H → F ∗ → 0. (1)

H is a torsion-free divisible abelian group (written addi-

tively), F is an algebraically closed field, and exp is the

map from H to F ∗.

We can code this sequence as a structure:

(H, +, E, S),

where E(h1, h2) iff exp(h1) = exp(h2)

and

S(h1, h2, h3) iff exp(h1) + exp(h2) = exp(h3).

33

AXIOMATIZING COVERS

Theorem. There is an Lω1,ω-sentence Σ such that there

is a 1-1 correspondence between models of Σ and sequences

1.

Lω1,ω comes in to say the kernel is 1-generated (and there-

fore countable).

Theorem. Σ is quasiminimal excellent.

34

THUMBTACK LEMMA

Definition 12 A multiplicatively closed divisible subgroup

associated with a ∈ C∗, is a choice of a multiplicative

subgroup isomorphic to Q containing a .

Theorem 13 (Zilber’s thumbtack lemma)

For any b1, . . . b` ⊂ C∗, there exists an m such that

b1m1 ∈ bQ1 , . . . b

1m` ∈ bQ` ⊂ C∗, determine the isomorphism

type of bQ1 , . . . bQ` ⊂ C∗ over F .

35

CATEGORICITY IN Lω1,ω

Fundamental Fact I. To study categoricity in power of sen-

tences in Lω1,ω it suffices to study the class of atomic mod-

els of a complete first order theory – an atomic AEC.

Fundamental Fact II. Under weak CH, categoricity in an

uncountable power for an atomic class implies ω-stability

(for atomic types over models).

Fundamental Fact III. In an ω-stable atomic class, we may

assume that nonsplitting defines a dependence relation

which satisfies the properties of non-forking in an ω-stable

theory.

Proved by Shelah, 25 years ago, using work of Keisler and

the Lopez-Escobar theorem.

The we may assume is crucial. By expanding the lan-

guage we obtain stronger and stronger properties on the

class to be analyzed without changing the spectrum func-

tion.

36

INDEPENDENT SYSTEMS

Notation 14 An independent (λ, n)-system is a fam-

ily of models 〈Ms : s ⊂ n〉 such that:

1. Each Ms ∈ K has cardinality λ.

2. If s ⊆ t, Ms ≤ Mt.

3. For each s, As =⋃

t⊂s Mt is atomic.

4. For each s, Ms ^

As

Bs where Bs =⋃

t6⊂s Mt.

Definition 15 K satisfies the (λ, n)-existence prop-

erty if there is a primary (i.e. strictly constructible)

model over⋃

t⊂n Mt for every independent (λ, n)-system.

37

EXCELLENCE IN GENERAL

Definition 16 The atomic AEC K is excellent if

1. K is ω-stable;

2. For every n, K satisfies the (ℵ0, n)-existence prop-

erty.

38

CRUCIAL STEPS

Theorem 17 If K is excellent then

K has the (λ, n)-existence property for all n and λ.

.

Lessmann has shown that this last property is the ‘active

ingredient’ for deducing categoricity in all powers from

excellence.

Theorem 18 If for all µ < λ, the (µ, 2)-existence prop-

erty holds then for any model M of cardinality λ and

any a such that Ma is atomic, there is a primary model

N over Ma.

Corollary 19 K has arbitrarily large models.

39

QUASIMINIMALITY IN ATOMIC CLASSES

Definition 20 The type p over A ⊆ M ∈ K is big if

for any M ′ ⊇ A there exists an N ′ with M ′ ≤ N ′ and

with a realization of p in N ′ −M ′.

Definition 21 The type p ∈ Sat(A) is quasiminimal if

p is big and for any M containing A, p has a unique

extension to a type over M which is not realized in M .

Lemma 22 Let K be excellent. For any M ∈ K,

there is a c ∈ M and a formula φ(x, c) which is quasi-

minimal.

40

GEOMETRY AGAIN

Definition 23 Let c ∈ M ∈ K and suppose φ(x, c)

generates a quasiminimal type over M . For any el-

ementary extension N of M define cl on the set of

realizations of φ(x, c) in N by a ∈ cl(A) if tp(a/Ac) is

not big.

Theorem 24 This closure defines a homogeneous pre-

geometry on the quasiminimal set.

41

PARADISE REGAINED

Theorem 25 Suppose K is *-excellent. The following

are equivalent.

1. K is categorical in some uncountable cardinality.

2. K has no two cardinal models.

3. K is categorical in every uncountable cardinal.

Recall:

Theorem 26 [Shelah]

1. (For n < ω, 2ℵn < 2ℵn+1) A complete Lω1,ω-sentence

which has few models in ℵn for each n < ω is ex-

cellent.

2. (ZFC) An excellent class has models in every car-

dinality.

3. (ZFC) Suppose that φ is an excellent Lω1,ω-sentence.

If φ is categorical in one uncountable cardinal κ

then it is categorical in all uncountable cardinals.

1. Weak GCH and categoricity up ℵω implies excellence.

2. Categoricity up to ℵn does not suffice.

Corollary 27 Suppose K is *-excellent. If K is not

ℵ1 categorical, then K has at least n + 1 models of

cardinality ℵn for each n < ω.

42

Study (C, +, ·, exp)

Zilber’s program:

The most ambitious aim of the pseudo-analytic model pro-

gram is to realize (C, +, ·, exp) as a model of an Lω1,ω-

sentence discovered by the Hrushovski construction. This

program has two parts.

A. Expand (C, +, ·) by a unary function which behaves

like exponentiation using a Hrushovski-like dimension func-

tion. Prove some Lω1,ω-sentence Σ is categorical and has

quantifier elimination.

B. Prove (C, +, ·, exp) is a model of the sentence Σ found

in Objective A.

43

A: PSEUDO-EXPONENTIATION

K is the class of algebraically closed fields F equipped

with a unary function E such that:

1. E is a surjective map from F to F ∗, which a homomor-

phism between the additive and multiplicative group.

2. ker E = Z.

3. Schanuel’s conjecture holds for E:

If x1, . . . xn are linearly independent

td(x1, . . . xn, E(x1, . . . E(xn) ≥ n.

4. F ∈ K is strongly exponentially algebraically closed:

For any ‘suitable’ variety V defined over a finite C ⊂F , there is a generic over C realization of V in F .

44

A QUASIMINIMAL EXCELLENT CLASS

For X ⊂<ω A,

δA(X) = td(spanQX ∪ (exp(spanQX))− ld(spanQX).

Now define a combinatorial geometry.

Definition 28 1. For M ∈ K, A ⊆ M , A finitely

generated,

dM(A) = inf{δ(B) : A ⊂ B ⊆ M,B ∈ K0}.2. For A, b contained M , b ∈ cl(A) if dM(bA) = dM(A).

Extend to closures of infinite sets by imposing finite char-

acter.

This extends the Hrushovski construction because K is

axiomatized in Lω1,ω not first order.

45

DONE

F ∈ K is strongly exponentially algebraically closed if

for any ‘suitable’ variety V defined over a finite C ⊂ F ,

there is a generic over C realization of V in F .

Theorem 29 K Lω1,ω-axiomatizable and is quasimin-

imal excellent.

The members of K with countable closure are categor-

ical in all uncountable powers. This class is Lω1,ω(Q)-

axiomatizable.

This concludes the proof of Objective A.

46

AND TO DO??

Objective B is reduced to proving the hypotheses of the

following theorem.

Theorem 30 If the Schanuel conjecture holds in C

and if the strong exponential closure axioms hold in

C, then

1. (C, +, ·, exp) ∈ K.

2. (C, +, ·, exp) has the countable closure property.

47

REPRISE

I. Some natural mathematical structures are better axiom-

atized in non-elementary logics.

II. The ‘pure’ analysis, like stability theory in the 80’s, is

well-ahead of the applications. One can hope that this

general analysis will yield applications akin to the applica-

tions of orthogonality and geometric stability to Diophan-

tine analysis in the 90’s.

III. In any case, the model theoretic analysis stimulates

interesting mathematical questions.

MONOGRAPH: Categoricity www.math.uic.edu/ jbald-

win

48